PDA

View Full Version : First Listen: Spotify DJ, YouTube Radio Builder



Colorado Media Newsroom
February 28th, 2023, 11:21 AM
From Radio Insight:

https://i0.wp.com/radioinsight.com/wp-content/images/2019/02/spotify.png?resize=200%2C200&ssl=1(This story was updated on Feb. 27, to reflect my listening to Spotify’s DJ.)
Last Tuesday, YouTube Music announced the rollout of a feature called ?Radio Builder? (https://www.theverge.com/2023/2/21/23609228/youtube-music-radio-builder-custom-stations) that would let users create customized radio stations?selecting core artists, choosing hits vs. discovery, and adding other filters.
Less than 24 hours later, Spotify announced its new ?DJ? feature. (https://www.billboard.com/pro/spotify-dj-feature-ai-commentary/) Using AI and voiced by Xavier Jernigan, part of an earlier short-lived attempt at a ?morning show,? DJ will guide you through your own library as well as front-sell new music recommendations, in an effort to draw your attention to songs that you might otherwise skip over.
Both features are the sort of personalization that have seemed inevitable from the dawn of streaming audio. Although a decade ago, I thought my personalized DJ would probably be made possible by voice-tracking, not AI. And I hoped the feature would come from a broadcaster, since voice-tracking was cheap (or cheaper) and plentiful.*
The YouTube feature was available when I went looking for it on Wednesday night, Feb. 22. When I first wrote this article last Thursday, I didn’t have Spotify’s DJ feature and neither did anybody I knew. Most of the early press coverage was based on Spotify’s demo (https://newsroom.spotify.com/2023-02-22/spotify-debuts-a-new-ai-dj-right-in-your-pocket/) and still is as of Monday (27). Most of the tweets on DJ are still of those announcement stories, although on Saturday, there was a negative early review from Wired (https://www.wired.com/story/spotify-ai-dj/) circulating as well. In speaking to radio people this week, the far more earth-shaking story has been Futuri’s announcement of “RadioGPT,” (https://radioinsight.com/headlines/249038/futuri-launches-radiogpt-to-bring-ai-created-audio-to-radio/) a product offering AI-generated topical jock breaks.
By Saturday morning, I was finally able to take a First Listen, thanks to two broadcasters who did have access to DJ: veteran programmer Rayne and Westwood One personality Bernie Mack, both of whom were kind enough to give me some time on the weekend to demo how the service was working for them.
On its own volition, “DJ X” is first heard when you activate the service and then seems to come back every 4-5 songs. (Rayne, the day before, had listened for a while and only gotten an intro break.) If you press the blue DJ button, he returns to set up the next segment, usually beginning with a comment like “okay, let’s change it up” or “I got you” before front-selling the next set. Some breaks were in the cold, but both Mack and Rayne noticed that “DJ X” did sometimes talk up intros and hit the post.
Mack’s music was based on a mix of his songs, his wife’s library, and, in one case, because there was a newborn at home, ambient noise music (which DJ X did front-sell in one case as “here’s where fans of ‘water’ are spending their time”). A lot of DJ’s content was reminiscent of Spotify’s year-end Wrapped analysis of user listening brought to life. “We’re switching it up to some of your old summer jams. Here’s Billie Eilish kicking it off,” said one break. A lot of the content, Mack noted, was “that was, this is.” There was also a set of David Bowie songs introed as “now a bunch of songs from an artist that you haven’t listened to lately. Consider this a reminder.”
Rayne had been using Spotify to listen mostly to a yacht rock playlist. “We’re going to kick it off this Saturday with Dan Fogelberg and those kind of vibes,” said DJ X before launching into “Hard To Say.” A few songs later, DJ X said “we’re going to keep it going with some folk acoustic, all for you” before playing Carly Simon’s “You’re So Vain.” Later, there was “I see you’ve had Ambrosia in rotation, so let’s go with a mix of their songs” and, again, “here’s an artist you used to listen to a lot” (into Firefall this time).
One of the stated goals of the new feature was to give extra emphasis to songs featured on current Spotify playlists. Mack got an Editor’s Pick recommendation for a new Hozier song. Rayne was getting recommendations for new Hip-Hop, including a new Lil Uzi Vert and a deeper Metro Boomin’ song, even though there was almost no Hip-Hop in his feed. Rayne also got a set of “top songs across the country” that featured NF, Miley Cyrus, Morgan Wallen, Bailey Zimmerman, and Coi Leray.
I was impressed, as were the users I talked to, with how non-robotic DJ X sounded. There were a number of times where he repeated himself, especially when Rayne kept requesting new content. (One heavily recurring theme was music that was “just your vibe.”) Without being told that there was AI involved in generating the content, it sounded more like Jernigan had cut a (relatively) large number of voice-tracks which AI was then assembling–e.g., matching generic lines with specific artist names.
At first blush, what I?d like from Spotify is a hosted version of Today?s Top Hits (or my own CHR playlist (https://open.spotify.com/playlist/0Hg1TtXOIiyGuONc6rFKQ0?si=ec22f2c69f9847e0)). I would prefer those hosts to be the seven best live DJs in the world presiding over the shared experience that TTH has become. At the outset, Spotify DJ is by design more concierge than companion, although horror movie fans will notice that he knows what you did last summer and reminds you of it a lot.
I am curious what Spotify DJ will make of my own*oddball taste and widely varying weekly usage. For years, that has often confounded any streamer?s ability to recommend music for me, although when I can listen to Spotify?s Discover Weekly, I will often be introduced to 2-3 songs out of thirty that I want to go back to. That’s not a bad batting average for me.
RADIO BUILDER
Unlike Spotify DJ, YouTube’s Radio Builder was available immediately when I went looking for it. https://i0.wp.com/radioinsight.com/wp-content/images/2023/02/ytmusic.jpg?resize=200%2C200&ssl=1YouTube Music would only let me pick from a selection of artists beginning with the most recent ones I had watched on YouTube. There was no way to enter artists from scratch. There were plenty of artists to choose from?more than a hundred, some that I barely remembered having watched, and mostly older since YouTube has not been my go-to for new music listening.*
I went for a mostly ?60s-through-?80s station. My seed artists ranged from Aretha Franklin to April Wine. Stevie Nicks to the Sweet to the SOS Band. The handful of newer artists I was able to choose included Mary J. Blige, Paramore, Taylor Swift, and Rihanna.*
I was then given the opportunity to choose the amount of artist variety (high), familiarity level (I went for the ?blend? between ?familiar? and ?discover?) and filters (I chose only ?upbeat,? but could have also gone for ?popular? or ?deep cuts?; ?pump-up? or ?chill.? Once the playlist was generated, there was the ability to hit ?tune? and revise artists or settings. To keep the playlist, I had to add it to the library. Otherwise, I was starting over from scratch.*
The hit rate was pretty good. Out of the first nearly 30 songs, about 60% are songs I would have enjoyed encountering on the radio; 20% are songs that I?m indifferent to, but probably would have left on, and 20% that seemed only superficially related to songs I liked (in the way that choosing ?Mickey? has always seemed to prompt ?Love Shack?).*
Here?s the beginning of the playlist that YouTube Music devised for me.


Four Tops, ?I Can?t Help Myself?
Rumours of Fleetwood Mac, ?Everywhere??performed by a tribute band; prompted by my watching the Niall Horan/Anne-Marie charity cover of ?Everywhere? last year
Miracles, ?You Really Got a Hold On Me?
Tavares, ?Heaven Must Be Missing An Angel?
Coolio, ?Gangsta?s Paradise?
Bobby Brown, ?Every Little Step?
John Travolta & Olivia Newton-John, ?You?re the One That I Want?
Creedence Clearwater Revival, ?Have You Ever Seen the Rain?
Cherrelle & Alexander O?Neal, ?Saturday Love?
Monkees, ?A Little Bit Me, A Little Bit You?
Sweet, ?Blockbuster?
Traveling Wilburys, ?End of the Line?
Whitney Houston, ?How Will I Know?
Marvelettes, ?Please Mr. Postman?
Billy Ocean, ?Love Really Hurts Without You?
Beyonc?, ?Super Bowl 2013 Halftime Show??13:15 worth of it
Pasadenas, ?I?m Doing Fine Now??one of several titles, including the Sweet and Billy Ocean songs, that were UK but not US hits
Fat Joe f/Ashanti, ?What?s Luv?*
New Edition, ?Mr. Telephone Man?
Rumours of Fleetwood Mac, ?Little Lies?
Steve Winwood, ?Higher Love?
Whispers, ?And the Beat Goes On?
Isley Brothers, ?This Old Heart of Mine (Is Weak for You)?
Jimmy Ruffin, ?What Becomes of the Brokenhearted?
Bucks Fizz, ?Making Your Mind Up??1981 UK hit/Eurovision winner
Kleeer, ?Intimate Connection??minor mid-?80s R&B hit, much further down was ?Teardrops? by Womack & Womack (another song that is a UK smash/US obscurity)
Gerry Rafferty, ?Baker Street?

Throughout my “future of radio” discussions, I’ve maintained that there will always be a demand for the things that radio does, even if it’s not necessarily on an FM over-the-air radio station. I’ve been surprised that streamers haven’t rushed to fill the void, but Amazon’s launch of Amp and Spotify’s second attempt at a radio experience are moves in that direction that broadcasters need to take seriously.



more (https://radioinsight.com/blogs/249036/first-listen-spotify-dj-youtube-radio-builder/)